Initial Management

14. In patients with pTBI, what are the benefits and harms of pre-hospital and ED resuscitation strategies (e.g., correction of coagulopathy, cardiopulmonary resuscitation, airway maneuvers, oxygen administration, cervical spine immobilization, specific fluid administration including hypertonic saline, mannitol administration, pain management, hyperventilation, posture), including the prevention of secondary injury (e.g., cerebral edema, diffuse axonal injury)?
15. In patients with pTBI, what patient and scene-related factors indicate a definitive airway is needed compared with no or non-invasive ventilation?
16. In patients with pTBI who require intubation, what are the benefits and harms of sedation and analgesia?
17. In patients with pTBI, what are the benefits and harms of on-scene wound care of head injury?
18. In patients with pTBI, what are the benefits and harms of early initiation (prehospital and ED) of antibiotic prophylaxis?
19. In patients with pTBI, what are the benefits and harms of treatment in the prehospital and ED with tranexamic acid, antiseizure prophylaxis, and pressor support?
20. In patients with pTBI, what are the benefits and harms of routine evaluation with thromboelastography and correction of coagulopathies?
21. In patients with pTBI, what are the benefits and harms of a 24-hour or more delay in evaluation by a neurosurgeon?
22. How should treatment be modiï¬ed if field care is prolonged (>24 hours)?
23. Do the effects of treatment for pTBI differ based on patient characteristics (e.g., age, gender, military status) or injury characteristics (e.g., mechanism of injury, location of injury)?

Critical Care Key Questions:

Intracranial Pressure

24. In patients with pTBI, what are the benefits and harms of ICP monitoring?
25. In patients with pTBI, what blood pressure thresholds should be used for ICP and cerebral perfusion pressure?
26. In patients with pTBI, what are the benefits and harms of treating intracranial hypertension (e.g., hyperosmolar therapy, lumbar drainage, ventilation therapies) in the critical care (i.e., intensive care unit) setting?

Prophylaxis

27. In patients with pTBI, what are the benefits and harms of seizure prophylaxis, either initiated or continued, in the critical care setting?
28. In patients with pTBI, what are the benefits and harms of antibiotic prophylaxis, either initiated or continued, in the critical care setting?
29. In patients with pTBI, what are the benefits and harms of deep vein thrombosis prophylaxis?

Other

30. In patients with pTBI, what are the benefits and harms of screening for delayed vascular complications?
31. In patients with head injury, when should organ donation occur?
32. In patients with pTBI, what are the benefits and harms of rehabilitation (e.g., physical, speech, occupational therapy) and when should rehabilitation be started?

Surgery Key Questions:

Timing of Surgery

33. In patients with pTBI who are neurologically stable, what are the benefits and harms of delaying surgery for better operative conditions?
34. In patients with pTBI, what are the benefits and harms of urgent surgical decompression/craniectomy vs a period of observation prior to decompression?

Type of Debridement

35. In patients with pTBI, what are the benefits and harms of local debridement and closure?
36. In patients with pTBI what are the benefits and harms of retained bone or metal fragments, or removal of retained/embedded objects (e.g., blade)?
37. In patients with pTBI, what are the benefits and harms of radical (brain debridement, deep fragment removal) vs conservative debridement (local irrigation and closure)?

Prevention of CSF leaks

38. In patients with pTBI, what are the benefits and harms of water-tight closure of the dura, with or without graft, to prevent cerebral spinal fluid (CSF) leak in the setting of exposure of cisterns or the ventricular system?
39. In patients with pTBI, what are the benefits and harms of CSF diversion?

Cranioplasty/facial surgery

40. In patients with pTBI, what are the benefits and harms of autologous cranioplasty vs use of alloplastic implants (i.e., titanium, methyl methacrylate, polyetheretherketone)?
41. In patients with pTBI, what are the benefits and harms of early vs delayed cranioplasty?
42. In patients with pTBI and facial injury, what are the benefits and harms of a multidisciplinary approach to pTBI surgery/management?

Traumatic cerebrovascular injury

43. In patients with pTBI, what are the benefits and harms of endovascular coil occlusion to reduce risk of rupture, in the setting of a traumatic aneurysm?
44. In patients with pTBI, what are the benefits and harms of surgical or endovascular management of venous sinus injuries to prevent venous infarction?
45. In patients with pTBI, what are the benefits and harms of obtaining intraoperative cultures to determine appropriate antibiotic treatment?
46. In patients with pTBI, what are the benefits and harms of surgery to reduce risk of rupture, in the setting of a traumatic aneurysm?

Search Strategies

We conducted electronic searches in Ovid MEDLINE , EMBASE, and Cochrane CENTRAL from inception to August 31, 2022 (Appendix A). We reviewed reference lists of systematic reviews for additional studies and the guidelines panel members provided references from their own hand searching. We also included studies from recent world conflicts that were published after the search dates as they became known to us.

Inclusion and Exclusion Criteria

Criteria were established a priori to determine eligibility for inclusion and exclusion of abstracts in accordance with the protocol developed by the guidelines panel. The criteria for inclusion and exclusion of studies for this systematic review are based on the KQs and PICOTS (Table 1).

For this review the population of interest was patients of any age with pTBI defined as a head injury that penetrated the dura and inflicted brain injury. This included studies of penetrating, perforating, and tangential injuries by any mechanism that addressed one or more of the KQs. Included interventions were screening for and diagnosis of pTBI, pre-hospital care, including on-scene wound care and resuscitation, medical management, including ICP monitoring and treatment, as well as seizure and antibiotic prophylaxis, and surgical management, which included management of CSF leaks, vascular injuries, and foreign body removal. We included all comparators and single arm studies with no comparator, if higher quality evidence was not available. Outcomes included accuracy of screening of diagnostic tools, mortality, neurological function, and morbidity and needed to be reported within 1 year of injury. We did not include outcomes of satisfaction, quality of life, sleep, or determining brain death. Because of the limited body of evidence it was decided a priori to include all study designs, including trials and observational studies that included case series and case reports when no other evidence was available. We did not include studies conducted in rehabilitation settings, but otherwise included all other settings (e.g., prehospital, battlefields, medical centers).

To ensure accuracy, all excluded abstracts were dual reviewed by two investigators. Each full-text article was independently reviewed for eligibility by two team members. All disagreements were resolved through consensus by two or three members of the systematic review team.

Study Selection

After full-text papers were retrieved and initial inclusion and exclusion screening was performed by the EPC, guideline panel members were assigned each KQ and reviewed the included articles to determine if the studies applied to their assigned KQ(s). This served as a double check of the evidence to help ensure no studies that provided evidence for any KQ were overlooked.

Risk of Bias Assessment of Individual Studies

Predefined criteria were used to assess the risk of bias (also referred to as quality or internal validity) for each individual included study, using criteria appropriate for the study design based on the Agency for Healthcare Research and Quality Methods guide, the Cochrane Back and Neck Group, and the U.S. Preventive Services Task Force (Appendix A). Each study was independently reviewed for risk of bias by two EPC members. Any disagreements were resolved through consensus. Based on the risk of bias assessment, included studies were rated as having "low," "moderate," or "high" risk of bias. Studies rated high risk of bias were not excluded a priori, but were considered to be less reliable than low or moderate risk of bias studies when synthesizing the evidence. Formal risk of bias assessment of case series and case reports was not done as these were determined to have high risk of bias due to study design.

Data Abstraction and Synthesis

For studies meeting inclusion criteria, evidence tables were constructed with the following data: study design, year, setting, country, sample size, patient characteristics (e.g., military vs civilian, age, race, mechanism of injury), and outcomes. All study data were verified for accuracy and completeness by a second team member. Additional details on data abstraction and synthesis see Appendix A.

Grading the Strength of the Body of Evidence

The strength of evidence (SOE) for each body of evidence was assessed as high, moderate, low, very low or insufficient, using the approach described in the Agency for Healthcare Research and Quality Methods Guide,35 based on study limitations, consistency, directness, precision, and reporting bias. These criteria were applied regardless of whether evidence was synthesized quantitatively or qualitatively. SOE ratings reflected our confidence or certainty in the findingss. For example, very low SOE means that we have very low confidence that future research will not change the findings. SOE was considered insufficient when evidence was lacking, sparse, or too conflicting such that we were unable to draw conclusions. SOE was initially assessed by one researcher and confirmed by a second. Descriptions of criteria and overall grades are described in full in Appendix A.

Development of Recommendations

The development of recommendations was the next step after the evidence was identified and synthesized and rated for risk of bias. Evidence-based recommendations are based on the quality of the body of evidence, applicability and generalizability. For topics where there is little or no research, recommendations were developed using rigorous Delphi consensus methodology.

Recommendations were assigned a level based on the quality of the body of evidence. We recognized five levels of recommendation based on the quality of the body of evidence:

• Level I recommendations are based on high quality of the body of evidence
• Level II recommendations are based on moderate quality of the body of evidence
• Level III recommendations are based on low quality of the body of evidence
• Level IV recommendations are based on very low quality of the body of evidence
• Level C: expert opinion recommendations based on consensus in the absence of research evidence

Delphi Process for Concensus

The rigorous consensus process employed in the generation of the SIBICC algorithms was replicated for this project which included overlapping leadership and participants. Blinded consensus voting was conducted to establish consensus statements complementing evidence-based recommendations. The same voting process was also used to ratify items for development of our consensus-based algorithms. As with the SIBICC algorithms,20,29 SurveyMonkey was used for voting prior to and subsequent to the in-person investigator meeting. At the in-person investigator meeting Electronic Meeting Services was hired to provide an independent consultant without medical expertise to perform the blinded voting using electronic keypads. Panelists ratified a threshold of 80% to declare consensus as having been achieved where at least 80% of the panel participate in the vote. This is the same threshold used for consensus in the recent SIBICC effort.20,29 Investigators met in Bethesda in a conference center generously provided by the Henry Jackson Foundation for five days on September 4-8, 2023 for intensive in-person discussion and work to advance all aspects of the project.

Overall Results

Results specific to each KQ are presented in relevant chapters. The systematic review used in our work is available as an online-only supplement (Supplemental Digital Content 5 [http://links.lww. com/NEU/F11]). Overall, a total of 6078 references from electronic database searches and manual searches were reviewed and 1923 full-text papers were evaluated for inclusion. We included a total of 125 studies in 135 publications, 29 were non-randomized studies, 85 single arm (case series) studies, and 11 case reports. A flow diagram of the search results and selection of studies is in Appendix B, a list of included studies is in Appendix C, and a list of excluded studies is in Appendix D. Evidence tables with study characteristics and detailed results are in Appendix E and risk of bias ratings for individual studies are in Appendix F.

Eighty-three studies (in 84 publications) were in civilian populations, and of these 48 were conducted in the United States; nine in South Africa; three in Brazil; two each in Canada, Colombia, Iraq, Israel, Lebanon, Switzerland, and Turkey; and one each in Afghanistan, China, Finland and Sweden, France, Iran, Italy, Palestine, South Korea, and Syria. Most civilian studies (57%, 32/56) reporting GCS scores reported the proportion of participants at specific scores or ranges of scores (e.g., GCS 3-8, GCS 9-12, and GCS 13-15) and did not report a mean or median score of the population. Fourteen studies reported mean admission GCS scores ranging from 6.7 to 14 and seven studies reported median admission GCS scores ranging from 3 to 10. One of these studies was rated low risk of bias, 23 were rated moderate risk of bias, while 58 were rated to be high risk.

Twenty-eight studies (in 36 publications) were in military populations, and of these three were conducted in the United States; three each in Iran, Turkey, and Vietnam; two each in Iraq, Syria, Croatia, and Ukraine; and one each in India, South Korea, South and North Korea, and Iraq and Iran. Similar to civilian studies, most military studies (61%, 11/18) reporting GCS scores reported the proportion of participants at specific scores or ranges of scores and did not report a mean or median score of the population. Seven studies reported mean GCS scores ranging from 7.5 to 10 and one study reported a median GCS score of 5. Two of these studies were rated moderate risk of bias, while 26 were rated high risk of bias.

Fourteen studies (in 15 publications) included both civilian and military populations; one was conducted in the United States; three in Croatia; two each in Lebanon, Israel, and Turkey; and one each in China, Egypt, Vietnam, and one Vietnam and the United States. One of these studies was rated moderate risk of bias, while 13 were rated high. As with the other populations, most of these studies (87%, 7/8) reporting GCS scores reported the proportion of participants at specific scores or ranges of scores and did not report a mean or median score of the population. One study reported a mean admission GCS score of 10 with a range from 3 to 15.

Two papers of the same population were conducted during the Lebanese conflict and the population was presumed civilian, but not specifically reported. This study is included with the civilian counts. Eighteen civilian studies reported race, White patients ranged from 22% to 85% of the population, Black patients ranged from 15% to 92%, Hispanic patients from 0% to 84%, Asian patients from 0% to 3%, and other unspecified or unknown races from 0.8% to 23%. The studies of military personnel and mixed civilian and military patients did not report race, additional characteristics of the patients included in these studies is reported in Table 2.

The studies of civilian populations addressed 25 of the KQs, while the studies of military populations addressed 11 KQs. Mortality was the outcome reported most often in studies, with infections, seizures, imaging findings, and prediction of survival with instruments or models, also often reported. Outcomes included and reported are different based on each KQ and sample sizes for the same study may be different for KQs based on the population included for each KQ. Results are reported in order of KQ, even though some KQs may relate to each other and have substantial overlap in included studies. Since we did not renumber the KQs when some were dropped, those that were dropped are noted in their location in the sequence of KQs as a reminder.

Numerous blinded surveys were conducted in conjunction with our Delphi consensus-process. Results of 5 online SurveyMonkey surveys conducted prior to our in-person meeting are available as online-only supplemental content (Supplemental Digital Content 6 [http://links. lww.com/NEU/F12], Supplemental Digital Content 7 [http://links. lww.com/NEU/F13], Supplemental Digital Content 8 [http://links. lww.com/NEU/F14], Supplemental Digital Content 9 [http://links. lww.com/NEU/F15], and Supplemental Digital Content 10 [http:// links.lww.com/NEU/F16]). Results of voting conducted during our in-person meeting are also provided (Supplemental Digital Content 11 [http://links.lww.com/NEU/F17] and Supplemental Digital Content 12 [http://links.lww.com/NEU/F18]). Two online SurveyMonkey surveys were conducted subsequent to the in-person meeting in conjunction with the review process (Supplemental Digital Content 13 [http://links. lww.com/NEU/F19] and Supplemental Digital Content 14 [http:// links.lww.com/NEU/F20]).

Summary of Key Findings

Twenty-six KQs were addressed in this review (Table 3): 125 studies provided evidence and another 80 studies provided contextual data from studies that did not meet criteria for inclusion but provided helpful information. Most of the evidence was of very low strength (meaning we have very low confidence in the certainty of effect) due to study design (e.g., case series, case reports) and/or few studies meeting inclusion criteria that answered the KQ (KQ31 and Appendix G). We did not identify any studies meeting inclusion criteria for 12 KQs (KQs 5, 6, 16, 17, 19, 22, 27, 32, 33, 39, 42, and 45). The highest quality evidence, rated moderate SOE (meaning we have moderate confidence that future research will not change the findingss), was identified for four KQs (KQs 4, 12, 13, and 38) that covered: digital subtraction angiography vs CTA, the relationship between bihemispheric injury in adult pTBI and mortality, the ability of the SPIN score18 to predict mortality, and the relationship between infection and CSF fistula, respectively. Preliminary work on KQ26 did not demonstrate that pTBI should be managed differently than other forms of TBI and we chose not to pursue work on KQ26 further. In the course of our work KQ20 was moved to ICU section, KQ 24-25 were combined and KQ30 was combined with KQ3.

Strengths and Limitations

Evidence for head injury care is generally accepted to be lacking compared with other fields of medicine, but this is especially true for penetrating brain injury. This guideline project thus employed liberal inclusion criteria needed to identify what evidence does exist. As a result, evidence for some KQs included case series and case reports. Appropriately, such evidence enabled only uniformly low level recommendations and consensus statements. In addition, the inclusion of case reports in this guideline highlights the need for higher quality future evidence focusing on interventions and the exploration of hypotheses. Another strength is that members of the guideline panel provided an additional review of the literature, effectively double checking that no relevant studies were overlooked by the systematic review team.

An important limitation of this work is the lack of high quality evidence for most KQs, requiring undesirable reliance on case series and case reports to fill in the gaps. Not surprisingly, no randomized trials were identified in this population. We identified moderate strength evidence for only four KQs, as noted above. The remaining evidence was considered low or very low strength or was insufficient to support any conclusions.

Another limitation is that most of the studies of pTBI were retrospective. In general, retrospectively collected data may be subject to recall and selection bias and represent a convenience sample that may not be representative of all patients who experience pTBI. Additionally, much of the evidence in pTBI comes from studies in military populations comprised of healthy young men. Older patients with significant co-morbidities may have distinct outcomes when treated similarly to military fighters and could potentially benefit from a distinct treatment strategy.

Gaps and Implications for Future Research

There was no evidence or very limited evidence for many of the KQs. This is largely due to the fact that pTBI is a rare injury that is often encountered in austere environments. This makes it difficult to study these patients or to prospectively accrue enough patients in a few months or years. It can also be difficult to ethically treat similarly-presenting patients too differently.

Given that most evidence in pTBI comes from case series, it is important that as much information from these series be gleaned as possible. In future studies with a mixed population (e.g., penetrating head injury vs closed head injury) patient outcomes should be reported separately by type of injury and in as much detail as possible to increase its usability. Future studies need to report results by the severity and/or mechanism of patient injury (e.g., Glasgow Outcome Scale score, missile injury vs blast injury). Additionally, studies employing multiple treatments or management strategies (e.g., fluid administration, use of blood products, airway maneuvers) need to also report patient outcomes by the specific treatments patients received. For example, many studies reported overall mortality as an outcome, but did not link mortality to any treatment or injury characteristic. Providing patient outcomes by population, injury, and intervention characteristics may enable findingss from similar studies (e.g., with similar populations, mechanism of injury, injury severity, and treatment strategies) to be pooled, which may result in stronger conclusions on the benefits, harms, or timing of any particular intervention or management strategy.

Summary and Conclusions

In order to provide useful clinical practice guidelines for pTBI we supplemented a traditional evidence-based approach with rigorous methods of identifying expert consensus. Few moderate strength conclusions related to specific management strategies for penetrating brain injury could be made. Detailed reporting of patient outcomes in future studies could advance the field by providing greater evidence for specific treatments by patient population, mechanism of injury, severity of injury, and specific interventions employed.